Trust intelligence infrastructure

Trust, computed.
Evidence attached.

GriffynX gives institutions a defensible basis for decisions about counterparties. We assess any individual, organisation or identifier from evidence gathered across independent sources, and return it with its confidence interval, the evidence that produced it, and the methodology version under which it was computed. Before onboarding, before authorisation, before disbursement.

Run a check on Discover Read the API contract

Both are available now. Neither requires an introductory call.

Sectors served

Seven with dedicated packs

80+

Evidence sources

One provider, one signal each

Endpoints published

Contract fixed and versioned

Decisions we make

The label remains your policy

one inquiry, end to end

The category

What trust intelligence is

Trust intelligence is the discipline of establishing, from evidence, how much confidence a decision can bear — and recording why it could bear that much.

It is a distinct function from verification, scoring, screening and detection. Those answer four narrow questions well. None answers the question an institution actually faces, which is not whether a document is valid or a transaction is unusual, but whether the party on the other side can be relied upon in this specific decision — and whether that judgement can be reconstructed afterwards.

The methodology

Category	The question it answers	What it establishes	What it is silent on
Identity verification	Is this document genuine, and does this face match it?	That a credential is valid.	Whether the holder should be trusted in this transaction.
Credit bureaux	How has this person repaid obligations before?	A borrowing history, where one exists.	Counterparties with no file, and every party that is not a borrower.
Screening and watchlists	Does this name appear on a list?	Membership of a published list.	Everything not yet on a list, which is where new conduct lives.
Fraud detection	Is this transaction anomalous against our own history?	Deviation from an internal baseline.	A counterparty with no history on your platform.
Trust intelligence	What can be established about this counterparty, from what evidence, with what confidence?	A computed assessment with its interval, its evidence chain and its methodology version.	The decision itself. That is your policy, and it stays yours.

Trust intelligence does not displace these. It is the layer that reconciles what they each return, together with evidence none of them collect, into one assessment that carries its own justification.

What we do

We produce one artefact: an assessment that carries its own justification.

An institution presents what a counterparty has offered it — an address, a number, a handle, a domain, a payment identifier, a device — together with the purpose of the enquiry and what is at stake. GriffynX gathers evidence from independent sources, reconciles it into an evidence graph, and computes an assessment from that graph alone.

The assessmentreturned on every call

score71: The assessment, on a fixed scale. Computed by one function from the admitted evidence, and by nothing else.
low · high58 · 84: The confidence interval. It widens when evidence is thin or in conflict, which is the information a single number destroys.
evidence12 signals · 0 contradictions: Counts by provenance class and by the banded trust of each source. Never a source name, never a raw payload.
findings[]statement · kind · trust_band · disposition: Each fact in plain language, with what sort of check produced it and whether it supports or weakens trust.
absences[]timeout · no_data · budget · refused: What was asked and did not answer, with the reason. An assessment that cannot state what it failed to learn overstates what it knows.
methodology_versionm1-graph: The version of the reasoning that produced this. An assessment issued in March can be reconstructed in October.
as_of2026-09-14T09:12:04Z: The moment the evidence was current. Assessments never change retroactively.
labelreview · policy v3: Your policy applied to the assessment, naming the policy version that applied it. The label is yours; we do not supply it.

The score is computed, never set

One function derives the score and the interval from admitted evidence. No model, analyst, customer, policy or engineer can set, adjust or tighten either. This is enforced in code and asserted in the test suite, not promised in a contract.

What did not answer is reported

Sources that timed out, refused, returned nothing or fell outside the budget are named in the assessment. Silent degradation is the failure mode that makes a vendor's output unusable in an examination.

Nothing raw crosses the boundary

No provider is named and no raw payload is returned. Each finding is a statement carrying its provenance class and the banded trust of its source; raw evidence is retained by hash for audit alone.

Assessments do not change retroactively

An issued assessment is a statement about what the evidence supported at a stated moment. When your risk appetite changes you publish a new policy version; the label changes and that change is itself recorded.

Industries

Twelve sectors. One engine, one contract, one evidentiary standard.

What differs between sectors is the decision being made, the artifact types presented, and what the institution is willing to tolerate in uncertainty. The reasoning does not differ, which is why a pack for a new sector is configuration rather than a release.

Banking and lending

The decision: Whether to open an account or extend credit against a thin file.
What existing tooling misses: A bureau reports borrowing history. It does not report that the email address on the application is three weeks old, that the device has submitted four applications under different names this month, or that the telephone number postdates the employment history it is offered to support.
What GriffynX adds: Correlation across identifiers the applicant supplied separately. A clean record on each one and an adverse pattern between them is the case this exists for.

Payments and fintech

The decision: Whether to authorise a transfer to a first-time payee, within the authorisation window.
What existing tooling misses: Rules detect what they were written to detect. A first-time payee at an unusual hour is either a fraud or a Tuesday, and the difference is not present in the transaction. It is in the conduct of the payee's identifiers elsewhere.
What GriffynX adds: A preliminary assessment inside the stated deadline, refined afterwards rather than withheld. UPI identifiers and DLT-registered SMS headers are first-class artifact types.

Digital assets

The decision: Whether to accept a deposit from, or release funds to, a wallet address.
What existing tooling misses: An address is sanctioned or it is not, and every exchange screens for that. The material question precedes it: whether the address, or the cluster around it, has been named in scam reporting, drained a contract, or appeared beside identifiers already refused.
What GriffynX adds: Sanctions screening and community reporting on a single call, with the address treated as an artifact that correlates to an email, a handle or a device rather than as a string in isolation.

Marketplaces and platforms

The decision: Whether to permit a seller to list, a driver to drive, or a host to host.
What existing tooling misses: Everything on the application is true. The applicant is real, the address exists, the documents are valid. What is not visible is that the identifiers reuse a network already behind three suspended accounts — a fact about your own platform's history joined to public evidence.
What GriffynX adds: Your suspensions are testimony. Submitted as observations, they weigh on future assessments, which no external vendor can do on your behalf because they do not hold them.

Dating and social platforms

The decision: Whether a new profile is a person, and whether it is the person it represents itself to be.
What existing tooling misses: Relationship fraud does not fail a document check, because the documents are genuine and belong to somebody. What discloses it is the shape of the account: a handle claimed across nine platforms in a week, photographs that predate the profile, a number in one jurisdiction and a device in another.
What GriffynX adds: Our broadest coverage sits on social handles, because identity is thinnest there. The assessment concerns the account, not the person behind it, and infers no protected characteristic.

Compliance and AML

The decision: Whether you can demonstrate, afterwards, that you looked and what you found.
What existing tooling misses: Screening tools answer the question asked on the day it was asked. What they rarely produce is what an examiner wants: which rule ran, on which version, against which evidence, and why the answer was what it was at that moment rather than today.
What GriffynX adds: Every call writes one audit record carrying the declared purpose, and every assessment names the methodology version that produced it. Risk bands are a policy you publish and version, so a label from March can be reconstructed in October.

Investigations

The decision: What occurred, who was involved, and what can be placed before a regulator or a court.
What existing tooling misses: The subject is a network, not a person. Several analysts touch it, scope grows as it proceeds, and at the end somebody must show their working to somebody who was not present.
What GriffynX adds: Cases carry an authorisation record, a budget, a legal hold and an append-only event log. Every expansion is one deliberate, costed hop rather than an open-ended crawl, and the export is an evidentiary bundle.

Also served

The same engine and the same artifact types. Sector-specific sources are in development, so these are scoped on a call rather than claimed here.

Insurance

Whether an applicant or a claimant is who the policy file says.

The same correlation across contact and device identifiers, ahead of underwriting and at first notice of loss.

Pack in development

Hiring and workforce

Whether a candidate or a contractor is authentic, before an offer or a first assignment.

Authenticity from public footprint alone, inferring nothing protected about the individual.

Pack in development

Real estate and rental

Whether a prospective tenant, buyer or listing agent is legitimate.

Domain age, infrastructure and contact-identifier conduct on the listing side as well as the applicant side.

Pack in development

Telecommunications

Whether to provision a line, a port-out or a bulk sender identity.

Number reputation, sender-identity registration and device correlation on the same assessment.

Pack in development

Travel and mobility

Whether a booking, a supplier or a counterparty in a high-value itinerary is genuine.

Domain and payment-handle assessment inside the booking window, with a preliminary answer on a deadline.

Pack in development

Full sector detail

Method

Five stages, each of which leaves a record.

Not a lookup and not a rule engine. Every fact is admitted with its provenance, every conclusion is derived from weighed evidence, and every stage writes to an append-only log that can be replayed.

01
Admit
Presented identifiers are normalised and typed against the registry. A purpose is mandatory. The stake fixes the tolerable interval and the evidentiary budget.
02
Gather
Independent sources are queried within that budget. One tool, one provider, one signal per call; no tool calls another or aggregates on its own authority.
03
Connect
Signals and the relationships sources reported become edges in an evidence graph. Typologies name the structures that matter — a mule chain, a synthetic identity.
04
Conclude
The confidence function derives the score and interval from that graph, stamped with the methodology version and the moment the evidence was current.
05
Retain
The assessment, the evidence hashes and any outcome you later report enter an append-only ledger. The next enquiry about the same entity does not start cold.

Provenance on every signal

An unprovenanced signal never reaches an assessment: where it came from, when, and the banded trust that source has earned.

Correlation is the product

A counterparty clean on each identifier separately and adverse in the relationships between them is the case this exists for.

Reproducible on demand

Methodology version and as-of timestamp on every assessment, append-only ledgers, and replay. An answer acted on in March can be reproduced in October.

Institutional memory

Outcomes you report weigh on subsequent assessments of your own book. The ledger is built only by the engine's writes; nothing is seeded into it.

Products

Two delivery surfaces, both live today.

The same kernel answers both. Discover is the evaluation surface, open without registration; the Trust API places the identical assessment inside your own decision flow, under your own policy.

Discover

Live · no registration

The evaluation surface at discover.griffynx.com. Submit an address, a number, a handle or a corporate domain and receive a written assessment: the verdict and its basis, each finding with its source class and confidence, the adverse indicators set against the supporting ones, recommended next steps, and an explicit statement of what could not be established.

Individuals and corporate entities
Assess a counterparty, or your own exposure
Structured intake, at most three questions
Unreachable sources named, never omitted

Open Discover What the report contains

discover.griffynx.com

I'm an individualI'm a companyX-Ray

Tell me what happened, or pin an email, phone number or social handle.

Someone on Instagram wants an advance for a puppy before I meet them.

Has this person asked you to move the conversation off Instagram, to WhatsApp or Telegram?

The trust verdictmixed · 51

9 checks ran6 found something3 still unknown

A disposable line, an account younger than the story it tells, and a handle seen on two other platforms this month. The reasons are below.

Message Discover…

I understand that GriffynX Discover provides informational insights based on publicly available and accessible information. Results may be incomplete, outdated, or inaccurate and should be independently verified.

Trust API

Live · v1

The integration surface at api.griffynx.com. A single call carries what the counterparty presented, the purpose of the enquiry, what is at stake and how long you can wait. It returns the assessment, the findings, the absences and your policy's label.

Inquiries live; 43 endpoints published as a fixed contract
Preliminary assessment inside a stated deadline, refined after
India-first artifact types alongside the international set
Keys shown once and stored hashed; every call audited

API documentation What is live, what is contract

POST /v1/inquiriesrequest

POST /v1/inquiries
Authorization: Bearer gx_live_…
X-Purpose: fraud_prevention

{
  "subject": {
    "presentations": [
      { "type": "PHONE", "value": "+91 98765 43210" },
      { "type": "EMAIL", "value": "a.sharma@example.in" }
    ]
  },
  "intent":  "onboarding_screen",
  "purpose": "fraud_prevention",
  "stake":   { "kind": "account", "reversible": false, "deadline_ms": 3000 },
  "render":  "none"
}

one assessment, evidence attachedresponse

200 OK

{
  "inquiry_id": "0f4c…",
  "status": "concluded",
  "assessment": {
    "score": 71, "low": 58, "high": 84,
    "preliminary": false,
    "methodology_version": "m1-graph",
    "as_of": "2026-09-14T09:12:04Z",
    "label": { "value": "review", "policy_version": 3 },
    "evidence": {
      "signals": 12, "failed_signals": 2, "contradictions": 0,
      "by_trust_band": { "high": 4, "medium": 6, "low": 2 }
    }
  },
  "findings": {
    "findings": [
      { "kind": "account_age", "about_type": "EMAIL",
        "statement": "The address has been seen in public records for under 60 days.",
        "trust_band": "medium", "disposition": "weakens" }
    ],
    "absences": [
      { "about_type": "PHONE", "kind": "budget",
        "reason": "A carrier lookup was not run at this tier." }
    ],
    "coverage": { "checks_run": 14, "checks_answered": 12, "checks_unanswered": 2 }
  }
}

Illustrative values in the shapes of the published contract. Every field above is in the OpenAPI document at api.griffynx.com.

X-Ray

Pilot · by request

Discover with the tier ceiling lifted: licensed sources, the relationships between what they return, and the full written assessment on every check. Five-day evaluations for fraud, compliance and investigation teams ahead of an integration.

Request an evaluation

The operating environment

An assessment you cannot explain is now a liability you pay for.

Supervisors on four continents require a specific reason for an adverse decision and treat an unexplainable model as model risk. Liability for authorised fraud is moving onto the institution: mandatory reimbursement in the United Kingdom, shared responsibility in Singapore, statutory penalties in Australia, and explainability named as a guiding principle by the Reserve Bank of India.

At the same time the evidence base is fragmenting. Identifiers are disposable and rotate faster than any list updates, and the counterparties most exposed are those with the thinnest documentary record.

Our position on this

85–95%

of anti-money-laundering alerts are false positives. Institutions absorb the review cost and still miss the material case.

Industry estimates, 2026

US$5

is the total cost to a financial institution of every US$1 lost to fraud — the majority of it in process, review and friction on legitimate customers.

LexisNexis True Cost of Fraud, 2025

70%

of firms lost clients to onboarding delays in 2025. The control intended to prevent loss is itself a source of it.

Fenergo, 2025

< 1%

of global illicit financial flows is ever seized or frozen, and most investigations begin without any record of what a prior case established.

FATF–INTERPOL–UNODC, 2025

Standards we hold ourselves to

Constraints enforced in code, each with a test behind it.

These are not commitments in a master services agreement. They are invariants: the build fails if any one of them is violated, which is the only form of assurance that survives a change of staff.

Purpose is mandatory

An enquiry without a registered purpose is refused. Every case carries an authorisation record and a budget; every analyst action is audited.

Demography never reaches the score

No source, typology, route or policy default keys on a protected or proxy attribute. Identical evidence yields an identical assessment.

Three schemas, three owners

A provider's payload stays inside its tool, the signal belongs to the kernel, and the public contract belongs to the gateway. None leaks into another.

Data classes govern retention

Retention is keyed to data class, regime and purpose. Class-2 raw material is not stored by default, and a legal hold makes deletion provable.

Jurisdiction is data

No country, language, currency, regulator or identifier format exists below the pack boundary. A new market is configuration, not a release.

Memory is earned, not acquired

The ledger is written only by the engine's own operations. Nothing is seeded, purchased or curated into it.

Evaluate it before you speak to us.

Both products are available without an introduction: run a check on Discover, or read the contract your engineering team would integrate against. A conversation is only necessary for a pilot or a key.

Run a check API documentation

Trust, computed.Evidence attached.

What trust intelligence is

We produce one artefact: an assessment that carries its own justification.

The score is computed, never set

What did not answer is reported

Nothing raw crosses the boundary

Assessments do not change retroactively

Twelve sectors. One engine, one contract, one evidentiary standard.

Also served

Five stages, each of which leaves a record.

Admit

Gather

Connect

Conclude

Retain

Provenance on every signal

Correlation is the product

Reproducible on demand

Institutional memory

Two delivery surfaces, both live today.

Discover

Trust API

X-Ray

An assessment you cannot explain is now a liability you pay for.

Constraints enforced in code, each with a test behind it.

Purpose is mandatory

Demography never reaches the score

Three schemas, three owners

Data classes govern retention

Jurisdiction is data

Memory is earned, not acquired

Evaluate it before you speak to us.

Trust, computed.
Evidence attached.